A multicore processor is a device that contains two or more independent computing components which are known as “cores or data-processing cores” that read and execute program instructions. Manufacturers typically integrate the data-processing cores onto a host-chip multiprocessor or CMP, or in a single chip package.
At the present time, a dual-core processors are known as processors that have two cores, a quad-core processor are those that contains four cores, a hexa-core processor contains processors that contain six cores, etc., and so on up to 64 cores. It can be assumed that the number of cores on a host chip may further increase in the future.
Presently, a common network topology for interconnection involves metal interconnects of the type used inside the host chip and includes the use of buses, rings, two-dimensional meshes, and crossbars. The architecture of the network topology of multicore interconnects may be rather complicated and its development continues.
With the increase of computing power realized on a chip, the use of traditional electrical interconnects becomes a limiting factor, i.e., preventing further scaling of modern multicore processors. A solution envisioned for future generation high-performance processors involves employment of optical interconnects. Replacing conventional electrical wires with optical waveguides helps to increase the overall bandwidth of data transmission, maintains channel crosstalk at a low value, and reduces the chip footprint allocated for interconnections.
Devices for optical interconnects between various electronic components mounted on the same or different PC boards of a computer, or between two or more independent chips, etc., are known in the art. The idea of optical interconnects per se is so obvious that it emerged much earlier than the appearance of multicore processors. In the past, the problem of optical interconnect has been solved in different ways.
For example, U.S. Pat. No. 5,832,147 issued Nov. 3, 1998 to Yeh, et al, discloses a holographic optical interconnect system and method for board-to-board and chip-to-chip communication interconnections. Each circuit board has at least an optically transparent substrate mate parallel to the circuit board and extending outside a circuit board holder. Each optically transparent substrate mate has parallel sides and carries at least two holographic optical elements. The first holographic optical elements on the first optically transparent substrate mate reflects at least a predetermined portion of a first light beam transmitted by a transmitter on a corresponding circuit board to another holographic optical element, which transmits a received light beam via free space outside the circuit board holder. On the other optically transparent substrate mate, two holographic optical elements are used to receive and direct at least part of the light beam received to a detector on a corresponding circuit board via free space within the circuit board holder or reflection within the optically transparent substrate mate.
U.S. Pat. No. H738 issued Feb. 6, 1990 to McManus, et al, discloses a switched hologram for reconfigurable optical interconnect. The device uses an array of optical switches that directs a set of optical beams toward any one of a selection of holograms, and each hologram when selected deflects the input beams toward a detector. The switches are arranged in rows, each optical switch includes a liquid crystal cell and a polarizing beam splitter, and each switch of said rows has an output face through which optical parallel beams can be projected to holographic spots of a hologram plate located adjacent to said output face and further from the holographic spots to detector means
U.S. Pat. No. 5,515,462 issued May 7, 1996 to Huang, et al, discloses an optical interconnect in optical packages using holograms for interconnection of two optical elements. Light radiation from the first optical component, e.g., an optical light fiber, is directed onto the plane surface of the second optical component, e.g., a passive waveguide. The light is directed onto the second optical component at a convenient angle such that it impinges on the plane surface of the second optical component at an incident angle. An angularly selective coupling means, e.g., a volume hologram, is located between the first and second optical components for coupling the maximum amount of the light at the incident angle.
U.S. Pat. No. 5,706,114 issued Jan. 6, 1998 to Erteza discloses an optical communication system that uses holographic optical elements to provide guided wave and nonguided communication, resulting in high bandwidth and high-connectivity optical communications. Holograms within holographic optical elements route optical signals between elements and between nodes connected to elements. Angular and wavelength multiplexing allow the elements to provide high connectivity. The combination of guided and nonguided communication allows polyhedral system geometries.
U.S. Pat. No. 4,838,630 issued Jun. 13, 1989 to Jannson, et al, discloses an optical interconnect that employs planar volume Bragg hologram technology in two dimensions and comprises a dichromated gelatin planar volume Bragg hologram disposed in a glass planar optical path. Multiplexed Bragg plane sets selectively diffract information-bearing light signals such as voice, image, or computer data signals in a very large-scale integration (VLSI) system from a laser diode or a light-emitting diode (LED) source coupled to the planar optical path toward high-speed photodiodes. The holographic planar optical interconnect can interconnect up to 1000 different signals between VLSI microelectronic components and systems.
However, all optical interconnects described above accomplish merely interconnection functions without any modulation or other data control processes and also do not match with the latest developments in the field of multicore processor technique in view of limitations with regard to chip footprint allocated for interconnections. Furthermore, in spite of the fact that all known optical interconnect devices described above possess essential advantages over traditional electrical interconnects, they did not find wide practical application in the industry because they do not match the current semiconductor production technology.